Skip to content
JW & Blood

Medical context

Research on treatment with and without transfusion. Each study is presented with its clinical setting, patient group and limitations. Updated .

Results from different conditions and study designs should not be treated as directly comparable. Historical studies also predate the 2026 policy changes. Current policy →

Not medical advice. This page is a summary of published literature, not medical advice. It cannot tell anyone what to do about their own care or their own child's, and it is not a substitute for a conversation with the clinicians treating them.

What the Society says about the medicine

“Patients, including children, who do not receive transfusions usually fare as well as or better than those who do accept transfusions.”
“Why Don't Jehovah's Witnesses Accept Blood Transfusions?”, jw.org, read 18 September 2026. The sentence carries footnote c, which cites The Journal of Thoracic and Cardiovascular Surgery, Volume 89, No. 6, p. 918, and Heart, Lung and Circulation, Volume 19, p. 658.
“Myth: Many Witnesses, including children, die each year as a result of refusing blood transfusions. Fact: This statement is totally unfounded. Surgeons regularly perform such complex procedures as heart operations, orthopedic surgery, and organ transplants without the use of blood transfusions.”
“Why Don't Jehovah's Witnesses Accept Blood Transfusions?”, jw.org, read 18 September 2026
“In any case, no one can say for certain that a patient will die because of refusing blood or will live because of accepting it.”
“Why Don't Jehovah's Witnesses Accept Blood Transfusions?”, jw.org, read 18 September 2026
“In many countries, any patient can now choose to avoid blood-transfusion risks, such as blood-borne diseases, immune-system reactions, and human errors.”
“Why Don't Jehovah's Witnesses Accept Blood Transfusions?”, jw.org, read 18 September 2026
“At one time, the medical community generally viewed strategies for avoiding transfusions, so-called bloodless medicine, as extreme, even suicidal, but this has changed in recent years.”
“Why Don't Jehovah's Witnesses Accept Blood Transfusions?”, jw.org, read 18 September 2026
“Citations of peer-reviewed articles from leading medical journals that present evidence in support of clinical strategies to preempt allogeneic blood transfusion in a variety of medical and surgical settings.”
“Medical Information for Clinicians”, jw.org Medical Library, read 18 September 2026
“Clinicians should therefore consult each Witness patient directly and ensure that the patient's individual decisions are clearly documented and respected. … Upon request, Hospital Liaison Committees will continue to assist these patients and the clinicians who care for them.”
“Important Adjustment Regarding Primary Blood Components”, jw.org Medical Library, 18 September 2026

Studies reporting comparable outcomes

Chiefly planned procedures, with haemoglobin optimised in advance and a team that took the case on.

  • Elective cardiac surgery

    Supports the claim

    Thirty-day mortality in a single surgeon's Witness practice fell across two decades even as the patients got older and sicker, which the authors attribute to pre-operative erythropoietin to a target haemoglobin of 14 g/dL, warm blood cardioplegia, the Cornell University protocol and fast-track extubation.

    Figures
    Group A mean age 51 years, group B 68 years. “Thirty days mortality was 3% in group A and 1% in group B despite greater operative risk factors, with more redo, and lower ejection fraction in group B.” An iterative procedure was performed in 13% of group B.
    Who was studied
    500 Jehovah's Witness patients at one French private hospital: 250 operated on 1991–2003 (group A) and 250 from 2003 to 2012 (group B). No non-Witness comparison group; the comparison is between the two eras.
    Check

  • Cardiac surgery, literature as a whole

    Supports the claim

    A review of the cardiac surgery literature found that most published comparisons report no significant difference between Witness and non-Witness patients, including for mortality, and concluded that a bloodless protocol “does not appear to significantly impact upon clinical outcomes”. The authors call for larger multi-centre cohorts.

    Figures
    Narrative review; no pooled effect size is reported. Literature searched from 2010 to 20 August 2020. Source material is “case reports, case series and comparative cohort studies”.
    Who was studied
    Published studies of cardiac surgery in Witness patients, 2010–2020. A narrative review, not a meta-analysis, so it inherits whatever selection the underlying single-centre reports carry.
    Check

  • Cranial and spinal neurosurgery

    Supports the claim

    Witness patients managed in a blood-management programme had outcomes, blood loss and operative time similar to matched controls, with few complications recorded.

    Figures
    68 Witness patients (23 men, 45 women, mean age 53 ± 12 years), 19 cranial and 49 spinal procedures. Median decrease in haemoglobin 2.1 g/dL; one patient had a postoperative haemoglobin below 7 g/dL. One return to theatre to revise a lumbar pedicle screw; one patient with postoperative seizures. “No cardiopulmonary complications, sepsis, pneumonia, or wound infection were observed.” Cell salvage was used in 27 cases and blood retransfused in 13.
    Who was studied
    68 Witness patients over five years at one US health system, matched to a concurrent cohort of non-Witness controls by sex, age and procedure.
    Check

  • Acute gastrointestinal bleeding

    Supports the claim

    Declining transfusion was not associated with higher 30-day mortality, but the Witness patients were more often admitted to intermediate care and needed more intensive bloodless management.

    Figures
    131 Witness patients matched to 131 controls. Thirty-day mortality OR 0.68 (95% CI 0.30–1.47; p = 0.327). Intermediate care admission OR 3.41 (95% CI 1.46–8.95; adjusted p = 0.028). No significant difference in ICU admission, vasopressor use, length of stay or 30-day readmission. “Both groups reached the same median nadir hemoglobin of 6.6 g/dL.”
    Who was studied
    Retrospective propensity-score-matched cohort across nine hospitals in one US health network, October 2014 to October 2024. Patients who declined transfusion throughout the index admission.
    Check

  • Autologous stem cell transplantation for myeloma

    Supports the claim

    Transplantation without transfusion support produced engraftment times, hospital stay and survival comparable to transfusion-supported transplantation at the same centre.

    Figures
    24 patients had bloodless autologous transplantation and 70 had transfusion-supported transplantation. Median progression-free survival 36 months versus 44 months (P = .277); median overall survival not reached in either group at a median follow-up of 59 months (P = .627). “There was no transplant-related mortality at the 100-day or 1-year mark in either group.”
    Who was studied
    Retrospective case-control study at one US cancer centre, August 2006 to August 2016. The authors note most transplant centres decline the procedure to these patients, so the 24 are a selected group.
    Check

Studies reporting greater risk

Chiefly bleeding that was not scheduled, and treatment that depends on transfusion support to be given at full dose.

  • Obstetric haemorrhage — United States

    Contradicts the claim

    Witness women delivering at one New York hospital died of obstetric haemorrhage at many times the rate of the hospital's other obstetric patients. The authors recommend that patients be counselled about the risk.

    Figures
    332 Witness women, 391 deliveries. Obstetric haemorrhage in 6%. Two maternal deaths, “for a rate of 512 maternal deaths per 100,000 live births versus 12 maternal deaths per 100,000 live births (risk ratio, 44; 95% CI, 9-211).” Erythropoietin “was associated with a nonsignificant increase in hematocrit level.”
    Who was studied
    Eleven years of deliveries by Witness women at Mount Sinai Medical Center, New York, compared with the same hospital's general obstetric population over the same interval. The confidence interval is wide because the comparison rests on two deaths.
    Check

  • Obstetric haemorrhage — Netherlands

    Contradicts the claim

    A nationwide enquiry found Witness women at several times the general population's risk of maternal death, and a far higher multiple of the risk of dying specifically from major obstetric haemorrhage.

    Figures
    “The MMR for Jehovah's witnesses was 68 per 100,000 live births. We found a risk of 14 per 1000 for Jehovah's witnesses to experience serious maternal morbidity because of obstetric haemorrhage while the risk for the total pregnant population was 4.5 per 1000.” The authors conclude a six times increased risk of maternal death, a 130 times increased risk of maternal death because of major obstetric haemorrhage, and a 3.1 times increased risk of serious maternal morbidity because of obstetric haemorrhage.
    Who was studied
    All maternal deaths in the Netherlands 1983–2006 and all cases of serious maternal morbidity 2004–2006, across tertiary centres, teaching hospitals and general hospitals — a national record rather than one hospital's.
    Check

  • Obstetric haemorrhage — United Kingdom

    Contradicts the claim

    Over fourteen years at one inner-city hospital most Witness deliveries were uncomplicated and postpartum anaemia was the commonest problem. There was one maternal death, and the authors' 65-fold figure is their own extrapolation from that single death, not a measured rate.

    Figures
    90 women, 116 deliveries: 24% caesarean section, 10% instrumental, 66% normal vaginal. Postpartum haemorrhage of more than 1,000 mL in 6%. Mean post-delivery haemoglobin 11.10 ± 1.15 g/dL versus mean pre-delivery 11.81 ± 1.62 g/dL (P > 0.05, paired t test). “The single maternal death occurred after caesarean hysterectomy, which when extrapolated, resulted in a 65-fold increased risk of maternal death compared to the national rate.”
    Who was studied
    Fourteen years of Witness obstetric care at one inner-city UK hospital. The 65-fold figure rests on one death in 116 deliveries and is presented by the authors as an extrapolation.
    Check

  • Massive haemorrhage at life-threatening haemoglobin

    Contradicts the claim

    A trial of a haemoglobin-based oxygen carrier used, as its comparison group, patients who had refused red cells on religious grounds. At the lowest red-cell haemoglobin levels, most of that group died within thirty days.

    Figures
    Historical control group: 300 surgical patients who refused red cells on religious grounds. Among those at a nadir red-cell haemoglobin of 3 g/dL or less, “the 30-day mortality was 25.0% (10/40 patients) compared with 64.5% (20/31 patients) in historical control patients at these RBC [Hb] levels.” The 40 trial patients had a mean nadir red-cell haemoglobin of 1.5 ± 0.7 g/dL with total haemoglobin maintained at a mean of 6.8 ± 1.2 g/dL by the infused product.
    Who was studied
    171 patients given the product in trauma and urgent surgery, against a historical — not concurrent, not randomised — control group of patients who declined red cells. The product is not licensed for routine use.
    Check

  • Acute myeloid leukaemia

    Contradicts the claim

    Chemotherapy doses had to be reduced to limit myelosuppression when transfusion support was refused. Remissions were fewer, none were durable, and the risk of death was an order of magnitude higher than in matched patients given full transfusion support. Most of the deaths not caused by the leukaemia were caused by anaemia.

    Figures
    Nine Witness patients treated 1998–2007. “Reduced dose induction chemotherapy … still led to complete remissions in three out of nine of JW patients but was associated with a high rate of relapse. No durable remission was achieved. The overall hazard ratio for death was 12.1 compared to a matched control group treated with full transfusion support. The predominant cause of non-AML mortality was severe anemia (four out of five early deaths) and uncontrollable bleeding (n = 1).”
    Who was studied
    Nine patients at one German university department, matched pairwise against controls who received full transfusion support. A very small series; its results should not be treated as a current prognosis for every patient.
    Check

  • Acute leukaemia, literature as a whole

    Contradicts the claim

    A review of acute leukaemia in adult Witnesses reports that these patients “are often denied intensive chemotherapy for fear they could not survive myeloablation without blood transfusion support”, and that treating them “is challenging and carries a higher morbidity and mortality”. The authors argue refusal should nonetheless not be an absolute bar to treatment.

    Figures
    Narrative review. No pooled figures, rates or effect sizes are reported in the abstract; the statement about higher morbidity and mortality is qualitative.
    Who was studied
    Published literature on adult Witnesses with acute leukaemia, reviewed rather than pooled.
    Check

Mixed findings and limited evidence

Findings that support part of the claim and undercut another part, and settings where the only published evidence is one or two cases.

  • Patient blood management as standard practice

    Mixed or limited evidence

    Avoiding unnecessary transfusion is now mainstream: an international consensus conference recommends detecting and managing anaemia before major elective surgery and sets restrictive transfusion thresholds. These are context-specific thresholds for considering transfusion, rather than a policy of refusing it in all circumstances.

    Figures
    17 PICO questions; 145 studies analysed, including 63 randomised clinical trials with 23,143 patients and 82 observational studies with more than 4 million patients. Two strong clinical recommendations on thresholds: haemoglobin below 7 g/dL for critically ill but clinically stable intensive care patients, and below 7.5 g/dL for patients undergoing cardiac surgery. Ten clinical and twelve research recommendations in total.
    Who was studied
    Systematic reviews and GRADE meta-analyses of adult red-cell transfusion, searched to January 2018, adjudicated at a consensus conference in Frankfurt in April 2018.
    Check

  • Surgery at low pre-operative haemoglobin

    Mixed or limited evidence

    Across nearly two thousand operations on patients who declined transfusion, overall 30-day mortality was low, and it rose steeply with lower pre-operative haemoglobin — more steeply in patients with cardiovascular disease.

    Figures
    1,958 patients. Thirty-day mortality 3.2% (95% CI 2.4–4.0) overall; 1.3% (0.8–2.0) with pre-operative haemoglobin 12 g/dL or greater; 33.3% (18.6–51.0) with pre-operative haemoglobin less than 6 g/dL. The interaction with cardiovascular disease was significant at p < 0.03, and the interaction between blood loss and low pre-operative haemoglobin at p < 0.001.
    Who was studied
    Retrospective cohort of patients aged 18 and over who underwent surgery and declined transfusion for religious reasons.
    Check

  • Severe anaemia after surgery

    Mixed or limited evidence

    In this surgical cohort, no deaths were observed at postoperative haemoglobin of 7.1–8.0 g/dL. Mortality was 34.4% in the 4.1–5.0 g/dL group. Lower haemoglobin was associated with higher odds of death.

    Figures
    Of 2,083 eligible patients, 300 had a postoperative haemoglobin of 8 g/dL or less. At 7.1–8.0 g/dL, 0 died (upper 95% CI 3.7%) and 9.4% (95% CI 4.4–17.0%) had a morbid event. At 4.1–5.0 g/dL, 34.4% (95% CI 18.6–53.2%) died and 57.7% (95% CI 36.9–76.6%) had a morbid event or died. Adjusted for age, cardiovascular disease and APACHE II score, “the odds of death … increased 2.5 times (95% CI, 1.9-3.2) for each gram decrease in Hb level.”
    Who was studied
    Consecutive patients aged 18 and over who declined transfusion for religious reasons and had surgery between 1981 and 1994, restricted to those with a postoperative haemoglobin of 8 g/dL or less.
    Check

  • Surgery generally, pooled case series

    Mixed or limited evidence

    The paper usually cited as the base rate for deaths under the doctrine argues that transfusion is over-valued, while reporting a non-zero rate of death from anaemia. Its abstract gives a range, not the single figure commonly quoted from it.

    Figures
    Data pooled from 16 reports. “Analysis of these data supports the concept that approximately 0.5% to 1.5% of such operations are complicated by anemia resulting in death.”
    Who was studied
    Series of Jehovah's Witness patients “who were not given transfusion for operations during which transfusion is typically given”. The author states that no controlled trial exists.
    Check

  • Trauma and major haemorrhage

    Mixed or limited evidence

    This Witness-specific trauma series is descriptive. It reports that treatments were applied inconsistently and calls for an international resource centre for clinicians facing haemorrhage where blood is not an option. It reports no comparative mortality, because there was no control group.

    Figures
    34 patients, 50% with major trauma. Anaemia was a clinical problem for 13 (38.2%), with haemoglobin reaching a nadir of 69.7 g/l (95% CI 56.7–82.7) on average 5.1 days (95% CI 2.5–7.7) after admission. Six (46.2%) received tranexamic acid, nine (29.2%) oral or intravenous iron, five (38.5%) erythropoietin. Three patients received packed red cells and two received haemoglobin-based oxygen carriers.
    Who was studied
    Retrospective review of self-identified Witness patients presenting to one Australian major trauma centre between January 2010 and January 2017.
    Check

  • Severe neonatal jaundice

    Mixed or limited evidence

    Two newborns with haemolytic disease reached the point where exchange transfusion was indicated and their parents refused. A drug that blocks bilirubin production halted the rise in both. In both cases legal action to compel the transfusion was already under way, and the drug was not licensed: it was released by a research laboratory with case-by-case approval.

    Figures
    Two infants. Case 1: born at 35 5/7 weeks, birth weight 2,790 g; bilirubin rose for 75 hours under three banks of lights and a biliblanket before a single 6 µmol/kg dose. Case 2: 38–39 weeks, 4,140 g; bilirubin fluctuated between 13.8 and 25.8 mg/dL with “suggestive clinical signs of possible bilirubin encephalopathy” before a single dose at 56 hours.
    Who was studied
    Two case reports, in 2001, of an agent used under Food and Drug Administration approval obtained for each infant. Not a cohort and not a comparison.
    Check

  • Sickle-cell disease: acute chest syndrome

    Mixed or limited evidence

    Transfusion is the mainstay of treatment for acute chest syndrome. The report cited here describes one patient who survived it without transfusion on maximal intensive-care support. One case cannot establish a rate either way.

    Figures
    A single patient, with a haemoglobin nadir of 3.1 g/dL, treated with mechanical ventilation, sedation, paralysis, erythropoiesis stimulation and limitation of blood draws, who survived.
    Who was studied
    One case report. This page does not include a cohort study of sickle-cell disease in patients refusing transfusion.
    Check

Risks of transfusion

Transfusion can cause adverse reactions. These studies report specific risks in the populations and periods described.

Estimates of deaths associated with refusal

The sources collected here do not establish a reliable worldwide death count. AJWRB’s figures are estimates based on clinical studies and population assumptions. The studies below document particular outcomes; they cannot by themselves establish the total number of deaths attributable to the doctrine.

Estimates in circulation — every one of these is an extrapolation, not a count

  • Estimate33,246 cumulative excess deaths worldwide, 1961–2016 — an average of about 594 a year, rising to about 1,220 in 2016.

    Published by
    AJWRB, applying an extrapolation by its medical adviser Osamu Muramoto (2001).
    How it is made
    Takes a mortality figure from Kitchens's 1993 pooled analysis of Witness surgery, rounds it down to a 1.0% excess-mortality assumption for conservatism, applies a US transfusion-utilisation rate of 15 per 1,000 population to the Witness publisher population, and multiplies across an average worldwide publisher population of 3,957,868 for 1961–2016. AJWRB states the years 1945–1960 are excluded and that the result is “an estimate rather than a documented count of observed deaths”. Note that Kitchens's abstract gives a range — “approximately 0.5% to 1.5% of such operations are complicated by anemia resulting in death” — rather than the single 1.4% figure the extrapolation starts from.
    Check

  • Estimate46,544 cumulative excess deaths, 1961–2016, and 1,708 in 2016.

    Published by
    AJWRB, as the upper end of its own range.
    How it is made
    The same calculation run with the unrounded 1.4% figure instead of the conservative 1.0%, which AJWRB says increases the result by approximately 40%. AJWRB presents this and the lower figure together as “a range of epidemiological extrapolations rather than precise mortality counts”.
    Check

  • EstimateAn annual excess mortality factor of about 0.015% of the publisher population.

    Published by
    AJWRB, from a second and independent extrapolation by Marvin Shilmer.
    How it is made
    Derived from a New Zealand study: twenty-one observed deaths against approximately two expected had the Witness cohort experienced the same mortality as matched transfused patients — about nineteen excess deaths over ten years, or 1.9 a year, set against an average of approximately 12,700 New Zealand publishers. AJWRB notes this converges with the Muramoto factor although the datasets differ.
    Check

  • Estimate“A thousand lives a year”, and in the text “an estimated 1,200 or more Jehovah's Witnesses dying annually because of the current blood policy”.

    Published by
    Open Minds Foundation, in a two-part article by Lee Elder, founder of AJWRB (2017).
    How it is made
    None stated in either part. The thousand appears in the title; part one is a personal history and gives no calculation, and part two states the 1,200 figure without showing its derivation. The number is consistent with AJWRB's 1,220 for 2016 and appears to rest on the same extrapolation, but the articles themselves do not say so.
    Check

Commentary

This site’s assessment of the sources above.

The studies describe different clinical situations. Some report good outcomes without transfusion, including in planned surgery and gastrointestinal bleeding. Others report higher mortality in obstetric haemorrhage or serious limits on treatment for leukaemia. Results from one setting cannot establish the safety of refusal in another.

Transfusion has risks, and avoiding unnecessary transfusion is part of patient blood management. That does not establish that refusing transfusion in every circumstance is safe. The studies here also differ in era, patient selection and available treatments.

The reported death estimates depend on assumptions about transfusion use and excess mortality. They should not be presented as known totals, or used to claim that a particular patient would certainly have survived with blood.